12-15-2014, 05:00 PM
Abnormal, I'm not paid to type online 
The project we're working on right now is a high speed (f/2) telescope with coma that is constant across the field (i.e sensor/film), normally coma is supposed to be an x^4 function with field.
The image with the bendy line I posted early is perhaps difficult to understand if you are not already familiar with RIM plots. The x axis is a scan across the pupil of the lens at the image plane. The y axis is defocus, or where the ray focuses with relation to the image plane. Spherical looks the way it does because the sign with which the ray strikes flips (and should flip) and be symmetric about the optical axis itself (center of the lens). If you think about it, the rays from the top edge should be a quantity, e.g 2mm defocused. The botttom has traveled in the "opposite" direction in the tangential axis, so the sign flips.
Spherical is caused by the angle of incidence not being constant across the lens surface, and it does not vary with field. Good 'ol snell's law tells you that the angle of refraction is tied to the index of the lens and the angle of incidcence. The curvature of the lens changes the AoI across the ray height whcih causes a focusing effect. Unfortunately, with a simple sphere we cannot account for the fact that near the edge of the lens the angle of incidence "blows up," aspheres were devised as a way to make the curve more gradual at the edge. A decade or so after the "invention" of the asphere people realized that you could use a flat lens and asphere equations to more or less only bend the edge of it, which will allow you to correct coma, at least at the edge of the image circle, absolutely.
Like this:
[sharedmedia=gallery:images:1316]
It's easier to manufacture that type of asphere than one that has a funky shape that needs to be molded, because even though it doesn't have a base curvature, you can grind it with a curvature (the one in that image was first ground to a radius of 157 with diameter 16.5mm) then polish it to the aspheric profile by hand and get fairly repeatable results. The problem with aspheres like that is that they must be made to a wider diameter than you need and then cut down, since the edge becomes very sharp, which is bad not only for mounting but for presenting unwanted effects in the final image that your design process did not predict.
When imaging a point on-axis the extra light, which may be regarded as flare, wouldn't be very visible - it would just look like spherical... and the 50/1.5 has plenty of that. I would imagine the initial prototypes were checked against an MTF chart, passed that evaluation, and then given to someone to make some "real world" pictures with, which were not examined closely, or they assumed this artifact (assuming it is present in all copies of the lens) was due to the object itself and not the lens directly.
When you design a lens in cad, you essentially are conscious of all the aberrations and you see them on the RIM plot.
[sharedmedia=gallery:images:1317]
You can have the cad software compute MTF, and spot size, which will tell you how the lens performs in a more practical sense. If your rms spot isn't on the order of +/- 1/2 pixel pitch, your design is either too poorly corrected, or "too highly" corrected and you can take elements or aspheres away and reduce size, weight, cost, etc.
Above is the rim plot set for an f/1.2 lens. At field height=0 degrees you see pure spherical. The geometry/design of the lens flattens the upper ray, and it becomes quite well behaved. If a ray stays along the x-axis, it is perfectly focused thus yielding a sharp, aberration free picture. At 2/3 field, this lens becomes dominated by coma. Towards the edge of the image, the lens returns to pure spherical, due to the asphere (this is the lens that contains the asphere I showed above). If you integrated the absolute value of these curves with respect to x, you would find that in total the edge of the pupil has less defocus than the center, indeed this lens is sharper off-axis than on-axis, because vignetting cuts off all of the poorly behaved rays. This is how some lenses, like nikon's supertelephotos, may perform better in the corner than the center.
Anyway, when you design you work from what's called a lens data table. It has a series of radii of curvature, thicknesses, and materials. You give the cad program things it can vary and give it constraints to drive the optimize to the solution you want. If you want to design a lens that is best for astrophotography, you try to nuke coma, which will leave you with spherical in most cases. This is how samyang seems to optimize much of the time, by trying to remove coma.
With most lenses, you look at 0, 7/10, and full field since that allows you to more or less see "the whole picture." Because everything is spherical, nothing weird happens and all aberrations follow their theoretical curves (such as coma being x^4 with field). When you introduce an asphere, you need more sample points because the asphere will cause odd behaviors, and you must be conscious of this or you will literally never see it coming.
It's always good to set the f number of the lens incrementally higher than the design spec to see what happens to rays you didn't previously expect to exist unless you have only very "nice" surfaces. My guess is that samyang did not do that, and thus you see an odd behavior.
Aspheres are the way of the future in optical design currently in the spotlight, but they present many challenges merely by existing. My research is into freeform optics, which requries no symmetry about any axis. You can observe many many strange effects, such as field constant coma, with freeform optics. Aspheres are used in freeform optics, they aer even the basis for it, but generally they have smooth shapes and are tested independently to the system and then have a baffle placed around the portion you do not want to use.
Unfortunately, aligning freeform optics is an absolute bitch. We spent about fourteen hours in the lab aligning the coma-constant telescope a couple weeks ago, only to discover that the piece of equipment we aligned it to was not functioning perfectly. One service call later, that peice of equipment has now been moved and all of our alignment is invalid and will have to be done again.
FWIW, panasonic's 20nm PV roughness is not too impressive. For freeform experiments we like to manufacture our parts to +/- 1.6nm of the ideal surface, which is the limit of what our polishing equipment can do. The average spherical lens for imaging is polished to between 1/20~1/50 wave for the HeNe laser in the equipment used to measure surface roughness. That's not to say panasonic's result isn't good - it's a very good surface, but it is not especially good, it is merely better than what most manufactures are producing in the photographic industry.
The project we're working on right now is a high speed (f/2) telescope with coma that is constant across the field (i.e sensor/film), normally coma is supposed to be an x^4 function with field.
The image with the bendy line I posted early is perhaps difficult to understand if you are not already familiar with RIM plots. The x axis is a scan across the pupil of the lens at the image plane. The y axis is defocus, or where the ray focuses with relation to the image plane. Spherical looks the way it does because the sign with which the ray strikes flips (and should flip) and be symmetric about the optical axis itself (center of the lens). If you think about it, the rays from the top edge should be a quantity, e.g 2mm defocused. The botttom has traveled in the "opposite" direction in the tangential axis, so the sign flips.
Spherical is caused by the angle of incidence not being constant across the lens surface, and it does not vary with field. Good 'ol snell's law tells you that the angle of refraction is tied to the index of the lens and the angle of incidcence. The curvature of the lens changes the AoI across the ray height whcih causes a focusing effect. Unfortunately, with a simple sphere we cannot account for the fact that near the edge of the lens the angle of incidence "blows up," aspheres were devised as a way to make the curve more gradual at the edge. A decade or so after the "invention" of the asphere people realized that you could use a flat lens and asphere equations to more or less only bend the edge of it, which will allow you to correct coma, at least at the edge of the image circle, absolutely.
Like this:
[sharedmedia=gallery:images:1316]
It's easier to manufacture that type of asphere than one that has a funky shape that needs to be molded, because even though it doesn't have a base curvature, you can grind it with a curvature (the one in that image was first ground to a radius of 157 with diameter 16.5mm) then polish it to the aspheric profile by hand and get fairly repeatable results. The problem with aspheres like that is that they must be made to a wider diameter than you need and then cut down, since the edge becomes very sharp, which is bad not only for mounting but for presenting unwanted effects in the final image that your design process did not predict.
When imaging a point on-axis the extra light, which may be regarded as flare, wouldn't be very visible - it would just look like spherical... and the 50/1.5 has plenty of that. I would imagine the initial prototypes were checked against an MTF chart, passed that evaluation, and then given to someone to make some "real world" pictures with, which were not examined closely, or they assumed this artifact (assuming it is present in all copies of the lens) was due to the object itself and not the lens directly.
When you design a lens in cad, you essentially are conscious of all the aberrations and you see them on the RIM plot.
[sharedmedia=gallery:images:1317]
You can have the cad software compute MTF, and spot size, which will tell you how the lens performs in a more practical sense. If your rms spot isn't on the order of +/- 1/2 pixel pitch, your design is either too poorly corrected, or "too highly" corrected and you can take elements or aspheres away and reduce size, weight, cost, etc.
Above is the rim plot set for an f/1.2 lens. At field height=0 degrees you see pure spherical. The geometry/design of the lens flattens the upper ray, and it becomes quite well behaved. If a ray stays along the x-axis, it is perfectly focused thus yielding a sharp, aberration free picture. At 2/3 field, this lens becomes dominated by coma. Towards the edge of the image, the lens returns to pure spherical, due to the asphere (this is the lens that contains the asphere I showed above). If you integrated the absolute value of these curves with respect to x, you would find that in total the edge of the pupil has less defocus than the center, indeed this lens is sharper off-axis than on-axis, because vignetting cuts off all of the poorly behaved rays. This is how some lenses, like nikon's supertelephotos, may perform better in the corner than the center.
Anyway, when you design you work from what's called a lens data table. It has a series of radii of curvature, thicknesses, and materials. You give the cad program things it can vary and give it constraints to drive the optimize to the solution you want. If you want to design a lens that is best for astrophotography, you try to nuke coma, which will leave you with spherical in most cases. This is how samyang seems to optimize much of the time, by trying to remove coma.
With most lenses, you look at 0, 7/10, and full field since that allows you to more or less see "the whole picture." Because everything is spherical, nothing weird happens and all aberrations follow their theoretical curves (such as coma being x^4 with field). When you introduce an asphere, you need more sample points because the asphere will cause odd behaviors, and you must be conscious of this or you will literally never see it coming.
It's always good to set the f number of the lens incrementally higher than the design spec to see what happens to rays you didn't previously expect to exist unless you have only very "nice" surfaces. My guess is that samyang did not do that, and thus you see an odd behavior.
Aspheres are the way of the future in optical design currently in the spotlight, but they present many challenges merely by existing. My research is into freeform optics, which requries no symmetry about any axis. You can observe many many strange effects, such as field constant coma, with freeform optics. Aspheres are used in freeform optics, they aer even the basis for it, but generally they have smooth shapes and are tested independently to the system and then have a baffle placed around the portion you do not want to use.
Unfortunately, aligning freeform optics is an absolute bitch. We spent about fourteen hours in the lab aligning the coma-constant telescope a couple weeks ago, only to discover that the piece of equipment we aligned it to was not functioning perfectly. One service call later, that peice of equipment has now been moved and all of our alignment is invalid and will have to be done again.
FWIW, panasonic's 20nm PV roughness is not too impressive. For freeform experiments we like to manufacture our parts to +/- 1.6nm of the ideal surface, which is the limit of what our polishing equipment can do. The average spherical lens for imaging is polished to between 1/20~1/50 wave for the HeNe laser in the equipment used to measure surface roughness. That's not to say panasonic's result isn't good - it's a very good surface, but it is not especially good, it is merely better than what most manufactures are producing in the photographic industry.